pytorch学习之tile()函数

tile(a1,a2,...,an)

通过复制原张量中的第i轴的向量并填充将第i轴向量个数扩大为原本的ai倍

a = torch.randn(2, 4)
b = a.tile(3, 3)

print(a)
print(a.shape)

# tensor([[-0.3954,  0.2900, -0.2817,  0.3923],
        [-1.6481,  0.6842, -0.7763,  0.7977]])
# torch.Size([2, 4])

print(b)
print(b.shape)

# tensor([[-0.3954,  0.2900, -0.2817,  0.3923, -0.3954,  0.2900, -0.2817,  0.3923,
#         -0.3954,  0.2900, -0.2817,  0.3923],
#        [-1.6481,  0.6842, -0.7763,  0.7977, -1.6481,  0.6842, -0.7763,  0.7977,
#         -1.6481,  0.6842, -0.7763,  0.7977],
#        [-0.3954,  0.2900, -0.2817,  0.3923, -0.3954,  0.2900, -0.2817,  0.3923,
#         -0.3954,  0.2900, -0.2817,  0.3923],
#        [-1.6481,  0.6842, -0.7763,  0.7977, -1.6481,  0.6842, -0.7763,  0.7977,
#         -1.6481,  0.6842, -0.7763,  0.7977],
#       [-0.3954,  0.2900, -0.2817,  0.3923, -0.3954,  0.2900, -0.2817,  0.3923,
#         -0.3954,  0.2900, -0.2817,  0.3923],
#       [-1.6481,  0.6842, -0.7763,  0.7977, -1.6481,  0.6842, -0.7763,  0.7977,
#         -1.6481,  0.6842, -0.7763,  0.7977]])

# torch.Size([6, 12])

ai必须时大于等于1的整数，否则会报错